OpenHands
Open-source AI software engineering agent that writes, edits, runs, and tests code autonomously in a sandboxed environment.
OpenHands
OpenHands (formerly OpenDevin) is a platform for AI software engineering agents. It gives an LLM a sandboxed Linux environment with a shell, code editor, browser, and file system — letting the agent autonomously complete entire software tasks from natural language instructions.
Key Features
- Full environment access: agent runs bash commands, edits files, browses the web, and calls APIs in an isolated Docker sandbox
- Multi-agent support: orchestrator + specialist agents (CodeAct, browsing, file editing) collaborate on complex tasks
- Any LLM backend: OpenAI, Anthropic, local Ollama, and any LiteLLM-compatible model
- SWE-bench leader: consistently top-ranked on the SWE-bench Verified benchmark for autonomous issue resolution
- GitHub integration: clone repos, open PRs, and comment on issues without leaving the agent loop
- Extensible skills: community-contributed microagents for specific domains (docs, testing, security)
Quick Start
# Requires Docker
docker pull ghcr.io/all-hands-ai/openhands:main
export WORKSPACE_BASE=$(pwd)/workspace
docker run -it --rm \
-e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.40-nikolaik \
-e LLM_MODEL="anthropic/claude-sonnet-4-5" \
-e LLM_API_KEY="$ANTHROPIC_API_KEY" \
-v $WORKSPACE_BASE:/opt/workspace_base \
-v /var/run/docker.sock:/var/run/docker.sock \
-p 3000:3000 \
ghcr.io/all-hands-ai/openhands:main
Open http://localhost:3000 and describe a coding task.
Install via ai-supply
npx ai-supply add openhands-ai-software-engineer
Curated mirror of the open-source OpenHands (MIT). Get it from the source.
Compromise signals — malicious or tampered code (leaked secrets, backdoors, a dropped executable) — reduce the score, and known dependency CVEs carry a bounded penalty (they warrant review but never QUARANTINE — update the dependency to clear). Other dangerous-by-capability traits are risk surface, expected for some capabilities. Every finding is mapped to its OWASP control below.
Findings mapped to the OWASP Top 10 for LLM Applications (2025) and the OWASP Machine Learning Security Top 10. Expand any flagged control for the exact findings — compromise reduces the score; expected/risk-surface do not, except a known CVE, which carries a small bounded penalty (high/critical → Review).
The same gate an agent runs before installing (POST /api/v1/trust/openhands-ai-software-engineer/check). Click a policy:
Consume OpenHands programmatically. Authenticate with an API key or session — see Authorize an agent.
# Agents: CHECK BEFORE YOU INSTALL (no auth) — score, grade, level, capability manifest
curl https://ai-supply.store/api/v1/trust/openhands-ai-software-engineer
# Gate against your org policy (returns { pass, violations })
curl -X POST https://ai-supply.store/api/v1/trust/openhands-ai-software-engineer/check \
-H "Content-Type: application/json" \
-d '{"minGrade":"B","denyPermissions":["shell"],"denyUnknownEgress":true}'
# CLI
npx ai-supply add openhands-ai-software-engineer
# REST (install → download)
curl -X POST https://ai-supply.store/api/v1/listings/openhands-ai-software-engineer/install \
-H "Authorization: Bearer $AIM_KEY"
# MCP tool
install_listing({ "slug": "openhands-ai-software-engineer" })OpenAPI spec →Curated mirror — latest upstream source. See the repository for tagged releases.